This study investigates the microvascular involvement of diabetes mellitus (DM) and non-proliferative diabetic retinopathy (NPDR) using machine learning methods, particularly focusing on non-parametric techniques. The research design involves a retrospective cohort study with optical coherence tomographic angiography (OCTA) images from healthy, DM no DR, mild DR, and moderate DR groups. The study aims to evaluate the importance of various microvascular parameters in different stages of DM and DR, using random forest (RF) classification and Shapley Additive Explanations (SHAP) to determine feature importance. Key findings include: - Large choriocapillaris flow deficits are the most important parameter for differentiating healthy eyes from DM no DR eyes. - Superficial microvasculature is important for distinguishing healthy eyes from DM no DR and mild DR eyes, but not for mild DR versus moderate DR. - Foveal avascular zone (FAZ) metrics are generally less affected but become more significant with worsening DR. - The study provides valuable insights into the complex relationship between microvascular changes and disease progression, facilitating the development of early detection methods and intervention strategies. The study highlights the utility of machine learning, particularly non-parametric methods, in understanding the intricate interactions between microvascular changes and diabetic retinopathy, offering a deeper understanding of the disease's progression and potential therapeutic targets.This study investigates the microvascular involvement of diabetes mellitus (DM) and non-proliferative diabetic retinopathy (NPDR) using machine learning methods, particularly focusing on non-parametric techniques. The research design involves a retrospective cohort study with optical coherence tomographic angiography (OCTA) images from healthy, DM no DR, mild DR, and moderate DR groups. The study aims to evaluate the importance of various microvascular parameters in different stages of DM and DR, using random forest (RF) classification and Shapley Additive Explanations (SHAP) to determine feature importance. Key findings include: - Large choriocapillaris flow deficits are the most important parameter for differentiating healthy eyes from DM no DR eyes. - Superficial microvasculature is important for distinguishing healthy eyes from DM no DR and mild DR eyes, but not for mild DR versus moderate DR. - Foveal avascular zone (FAZ) metrics are generally less affected but become more significant with worsening DR. - The study provides valuable insights into the complex relationship between microvascular changes and disease progression, facilitating the development of early detection methods and intervention strategies. The study highlights the utility of machine learning, particularly non-parametric methods, in understanding the intricate interactions between microvascular changes and diabetic retinopathy, offering a deeper understanding of the disease's progression and potential therapeutic targets.